Academic Communication

Rethinking Software for Non-volatile Processors on Energy Harvesting Embedded Systems2016-03-28

Source: cnu       Clicks:       Size: small     middle       big

Speaker: Dr. Chun Jason Xue (Computer science, city university of Hong Kong)

 

Time: Mar 29, 2015 9:30—11:00

 

Location: Conference room, Teaching Building, CNU Second North Campus.

 

Abstract: Energy harvesting is a favorable alternative to power future wearable devices, with its advantages in terms of size, weight and convenience. However, harvested energy is naturally unstable and thus introduces frequent program execution interruptions. To enable accumulative program execution, non-volatile processors demonstrate promising results. With the need of efficient backup and resumptions, non-volatile processor is different from traditional processors and necessitates new designs in both hardware and software. While a few works have been done in hardware level, little effort has been seen at the software level. In this talk, we explore software re-designs for non-volatile processors. First, two stack-oriented optimizations are presented to reduce the content to back up prior to power failure.  As a result, backup and resumption can be done faster with smaller required size of non-volatile memory. Then, based on the observation that the backup procedure may induce errors, a consistency-aware check-pointing scheme is proposed to locate and eliminate the potential errors. We believe that this is the beginning of a series of software re-designs, which are necessary for non-volatile processors, to enable its wide-spread application in future energy harvesting powered embedded systems. 

 

Speaker Bio: Dr. Chun Jason Xue is currently an Associate Professor at City University of Hong Kong Computer Science Department. His research interests include non-volatile memories, embedded and mobile systems. He is currently Associate Editor for ACM Transaction on Embedded Computing Systems. He was the TPC co-chair for LCTES 2015, TPC co-chair for ISVLSI 2016, and has served as TPC members in premiere conferences such as DAC, DATE, RTSS, RTAS, CODES, EMSOFT and ISLPED.